The term, steganography, comes from Greek roots, Image steganography and steganalysis Mayra Bachrach and Frank Y.

Image steganography and steganalysis Mayra Bachrach and Frank Y. Shih Image steganography is used to embed covert messages in the form of files, text, or other images in digital images. The intent is to transmit hidden information. Steganalysis is the process used to detect hidden messages in images. Although steganography is not a new discipline, it has become increasingly important in today s digital world where information is often and easily exchanged through the Internet, email, and other means using computers. The need for better methods and techniques which can be used both to embed hidden information in images and to detect messages hidden in images is driving new research in the area of steganography and steganalysis. This article surveys image steganography and steganalysis. The aim is to introduce the uninformed reader to image steganography and steganalysis. The key concepts behind image steganography and steganalysis are explained. The history and origin of steganography are outlined. Steganography is compared with watermarking in technique and intent. Details of how images are represented are explained. Methods and algorithms used to embed messages in images and to detect embedded messages are explored. Currently available steganography and steganalysis tools are explained. A demonstration of how to embed hidden information in an image using available steganography tools is provided. 2011 John Wiley & Sons, Inc. WIREs Comp Stat 2011 3 251 259 DOI: 10.1002/wics.152 INTRODUCTION The term, steganography, comes from Greek roots, which means covered writing. 1,2 The intent of steganography is to hide a message in a medium in such a way that no one but the intended recipient knows that the message is even there. Steganography is not a new discipline. Throughout history there has been the need to send hidden messages through covert means. The literature provides many examples of the rustic uses of steganography. In ancient times, slaves were used as messengers. The head of a slave was shaved and a message tattooed on it. Hair was allowed to grow back in order to hide the message. The slave was then sent to the intended recipient of the message who would have the hair shaved to reveal the message. 1 4 Music scores were used in the 1600s to hide messages. Each note in the music score would correspond to a letter of the message. Acrostic writings and invisible inks used during the world Correspondence to: shih@njit.edu Department of Computer Science, New Jersey Institute of Technology, Newark, NJ, USA DOI: 10.1002/wics.152 wars were also early examples of steganography. With the development of photography, tiny images containing secrets were sent out by pigeon post, or hidden under ears, nostrils and under finger nails of human messengers. 2 Today, computer technology is used to embed hidden messages in images using steganographic algorithms and techniques. The images are then transmitted electronically. There are several steganographic software systems available that can be used to easily embed messages in images. Some of this software is available license-free. The military and other government agencies are interested in steganography in order to be able to safely transmit information secretly. They are also interested in being able to detect secret messages transmitted by criminals, terrorists, and other unfriendly forces. It is suspected that Al-Qaeda made use of steganography during the planning of the World Trade Center attack. 5 Experts in the field have conducted studies to determine whether images modified by steganography are widely present on the Internet. 6 Businesses and corporations are also interested in steganography and steganalysis. Disgruntled or unethical employees can use steganography to transmit private corporate information through digital Volume 3, May/June 2011 2011 John Wiley & Sons, Inc. 251

www.wiley.com/wires/compstats communication. Companies want effective steganalytic tools that would help them detect unauthorized distribution of their private information and trade secrets. 5 With the advancement of technology and the advent of the Internet and personal computers, digital steganography and steganalysis have become topics of interest. Research in the area of steganography and steganalysis is driven by the need for protection of copyrights on digital works, detection of copyright infringement, the need for unobtrusive communication by military and intelligence agencies, and detection by law enforcement agencies of steganographic methods used in criminal activities. 2 STEGANOGRAPHY Using steganography hidden information in the form of files, text, or other images can be embedded in digital images. Digital Watermarking is used to embed information (watermark) in an image which can be used to identify and verify the owner and authorized users of an image. Steganography and digital watermarking are related disciplines. Although many of the concepts and techniques used in watermarking and steganography are similar, the two disciplines differ in purpose. The existence of the watermark is usually known to users of the image. The watermark may or may not be visible although it usually does not alter the appearance of the image. Both watermarking and steganography are used to embed information which cannot be easily detected or altered in an image. However, watermarking is used mainly to prevent the violation of copyrights by unauthorized users of an image, whereas steganography is used to exchange hidden information. In the case of watermarking, hidden information is embedded so that the image can be transmitted digitally while preventing copyright infringement and illegal distribution. In the case of steganography, the intent is to embed hidden information in order to transmit the information without detection. Steganalysis is the art and science used to determine whether undetectable messages have been embedded in images using steganography. The steganographer uses steganalysis to detect, extract, disable, or modify the message before it reaches the recipient. A steganographer a can take an active or passive approach to steganalysis. This is explained in the literature through the classic example of the prisoner s problem. Alice and Bob are prisoners planning an escape. Wendy is the warden. Alice and Bob are allowed to exchange messages freely through channels which allow Wendy to inspect the messages. Wendy can take two approaches. The passive approach is to examine all messages exchanged to see if any secrets are being exchanged. The active approach is for Wendy to alter all the messages exchanged between Bob and Alice, so that it is very difficult or impossible for Bob and Alice to extract the secret messages. In this classic example, Alice and Bob are using steganography to hide the secret information in their messages. Wendy is using steganalysis to detect and distort or remove the hidden information. 6,7 Many techniques, algorithms, and software have been developed to both embed hidden information in images (steganography) and to detect whether an image has an embedded hidden message (steganalysis). Image Representation Image files which are common on the Internet today are a very good medium for digital steganography. They are easily and frequently shared on the Internet through email, posting on sites, and other digital means. They are large in size which makes it easier to embed information in such a way that it is not noticed by the uninformed user. Steganographic techniques exploit the properties of images and are used to manipulate bits in the image and embed information. Computer images are represented as arrays of pixel s. A pixel is a point of an image. The numeric of each pixel is stored in 3 bytes (24 bits) and represents a digital color. Each of the 3 bytes defines the amount of light intensity in each of the primary colors: red, green, and blue (RGB) and can hold s from 0 to 255 (s which can be stored in 8 bits 2 8 ). For example, 255 for the red and 0 for the blue and green will render the color red. Other s close to 255 for the red component and 0s for the blue and green will also render a shade of red which may appear to be the same color to the human eye. 1,8 The large size of image files can be attributed to how pixels are represented. For example, a 24- bit image (3 bytes per pixel) 600 pixels wide and 600 pixels in height would consist of 600 600 24 bits (8,640,000 bits). Some of the pixels could be adjusted in to correspond to letters without any noticeable effect on the appearance of the image. Large image files are often compressed for faster transmission. There are two kinds of compression, lossless and lossy. Lossless compression maintains all the information in the image when compressed. It is used for GIF files among others. Lossy compression which is used for JPEG files may result in some loss of information. Steganography software employs different techniques for processing images based on their compression algorithm. 1,3 252 2011 John Wiley & Sons, Inc. Volume 3, May/June 2011

WIREs Computational Statistics Image steganography and steganalysis When selecting an image for steganography, both the image and color palette are considered. Images with large areas of solid color are more likely to show variations that can occur as a result of embedding messages. Embedded messages will be less likely to be noticed in gray-scale images and images with subtle color variations. Spatial Domain and Frequency Domain There are two classifications of image steganography methods: spatial domain b and frequency domain. c The techniques in the spatial domain embed the secret messages directly into the intensity s of the image pixels. In the frequency domain, images are first manipulated with algorithms and transforms, and then the messages are embedded in the image. The methods in the spatial domain are considered simplest but also more susceptible to steganalytic attacks, less robust. The spatial domain is sometimes called the image domain. d The frequency domain is also known as the transform domain. 7,e COMMON APPROACHES There are two common approaches for image steganography: Least Significant Bit (LSB) substitution and Algorithms and Transformations. 5 LSB substitution is simpler but less robust than steganographic methods which use algorithms and transformations. Least Significant Bit Substitution The LSB substitution involves embedding information in the LSB of a number of bytes of a carrier image. By overwriting the LSB of any pixel, information is embedded in the LSB of a byte representing an RGB color. The change resulting from modifying the LSB of scattered bytes results in such slight changes that it is not detected by the human eye. Overwriting any other bit, especially the most significant bit would result in a much more noticeable change and distortion of the image. LSB substitution is more effective on 24-bit images than on 8-bit images. LSB substitution is more effective with images that use lossless compression such as bmp and gif files. 1,9 LSB substitution is in the Spatial (image) domain. A clear example of the result of embedding the letter A in a 24-bit image is provided by Johnson and Jajodia (Ref 1, p. 28 29). In their example, 3 pixels (9 bytes) of an image contain the binary s: (00100111 11101001 11001000) (00100111 11001000 11101001) (11001000 00100111 11101001) The binary for the letter A is 10000011. Inserting the binary for the letter A in the 3 pixels results in: (00100111 11101000 11001000) (00100110 11001000 11101000) (11001000 00100111 11101001) The letter A was embedded into the LSB of each byte. But only the 3 bits underlined in the 9 bytes were modified as a result. All the other bits remained the same. LSB substitution is also used in watermarking. However, LSB substitution is more easily detected or made unusable by steganalytic tools than other methods of steganography. LSB Algorithm There are many variations of the LSB algorithm some of which are less susceptible to detection than others. The LSB algorithm may be easily used with gray-scale images. Each pixel in a gray-scale image consists of 8 bits. The first bit to the left is the most significant digit and the first bit on the right is the least significant digit. The pseudo code below can be used to explain the processing to embed a text message in a gray-scale image by replacing the LSB of each pixel: pic=cover image msg = secret message n = number of chars in msg for i = 1 to n get char from msg for each bit in char get a pixel from pic if the bit = 1 insert a 1 in the least significant bit of the pixel else insert a 0 in the least significant bit of the pixel replace the pixel in pic end for end for For a 24-bit image, the algorithm can embed more information per pixel. A 24-bit image utilizes 3 bytes, 24 bits to store the for each pixel. The first 8 bits in each pixel represent the color red, the second 8 bits represent the color green, and the last 8 bits represent the color blue. The pseudo code below can be used to explain a simple LSB insertion algorithm used with 24-bit images: Volume 3, May/June 2011 2011 John Wiley & Sons, Inc. 253

www.wiley.com/wires/compstats pic=cover image msg = secret message n = number of chars in msg for i = 1 to n Get char from msg for each 3 bits in char get a pixel from pic get the red of the pixel if the first bit = 1 insert a 1 in the least significant bit of the red else insert a 0 in the least significant bit of the red get the green of the pixel if the second bit = 1 insert a 1 in the least significant bit of the green else insert a 0 in the least significant bit of the green get the blue of the pixel if the third bit = 1 insert a 1 in the least significant bit of the blue else insert a 0 in the least significant bit of the blue replace the in pic end for end for The changes resulting from the LSB insertion algorithm applied to the LSB are not noticeable to the human eye. Even changes made to the second and third LSBs of each pixel are not visible to the human eye. In order to embed a larger message, information is sometimes hidden in the second and third bits of each pixel. Generally speaking, the LSB algorithm has a higher capacity than other embedding techniques. This means that a greater amount of information can be embedded per image. 5 The down side is that message embedded using LSB insertion can be easily destroyed by compressing, filtering, or cropping the image. 10 Steganographic techniques based on the LSB algorithm vary in complexity and robustness. The simple algorithm described above inserts the bits of the hidden message sequentially into the cover image. f As a result, it is easy to detect and extract the message. Some variations of the LSB insertion algorithm insert the bits randomly into the cover image based on a stego key. g For example, one variation of LSB insertion uses the random pixel manipulation technique to insert the message into random pixels in the cover image. The random pixel manipulation technique utilizes a stego key. The stego key provides a seed for a random number generator. Using the seed, random pixels in the image are selected for embedding the message. The stego key is then used to extract the message by using the same seed number to generate the random pixels where the data has been inserted. 10 Although inserting the message in random pixels makes detection and extraction of the hidden message less likely, the hidden message can still be destroyed by compression and other image manipulation such as filtering or cropping. 1 Some of the steganographic tools available modify the color palettes of the image in order to make detection and extraction of the message less likely. Examples of tools using these techniques, such as S-Tools and EZStego will be described later on in this article. Algorithms and Transformations Steganographic methods utilizing algorithms and transformations are more complex but also more robust. They are in the frequency (transform) domain. Some of these methods utilize the Discrete Fourier Transform (DFT), the Discrete Cosine Transform (DCT), the Discrete Wavelet Transform (DWT), and the Genetic Algorithm (GA). 1,3,5,11 Steganographic methods in the frequency domain apply transformations such as the DFT or DCT to an image and then embed information in the bits of the coefficients obtained from the transforms. 12 Discrete Fourier Transform The DFT is based on the Fourier series used to represent the continuous time periodic signal. 5 The Fourier series which is the basis for the DFT states that any periodic function can be expressed as the sum of sines and/or cosines of different frequencies multiplied by a different coefficient. 13 The DFT decomposes an image into its sine and cosine function. Using the inverse DFT an image which has been transformed using DFT can be transformed back into its spatial domain equivalent image. Steganographic methods which use the DFT embed information by modifying the bits of the resulting DFT coefficients. Because the calculation of the DFT for an image is processing intensive, the Fast Fourier Transform (FFT) is used to derive the results. 5 Languages such as MATLAB which are used for image processing provide functions that can be used to derive the DFT of an image. 254 2011 John Wiley & Sons, Inc. Volume 3, May/June 2011

WIREs Computational Statistics Image steganography and steganalysis The two-dimensional Fourier Transform 14 takes a two-dimensional image matrix, f (x, y) as input and outputs another two-dimensional matrix, F(u, v). The calculations using Fourier Transform are resource intensive. Languages such as MATLAB rich in image processing tools are used to derive the transforms. In particular, MATLAB has the Fast Fourier Transform (fft, fft2) and Inverse Fast Fourier Transform (ifft, ifft2) functions which can be used in processing images. Discrete Cosine Transform The DCT 5 is used to compress jpeg image files. Using steganographic techniques, information can be hidden in jpeg images during the compression process. Using the DCT, blocks of 8 8 pixels of a jpeg image are transformed into 64 DCT coefficients. The DCT coefficients are quantized using a 64-element quantization table. The LSBs of the quantized DCT coefficients are used to embed the hidden information. 4,5,11 The image can be transformed back into its spatial domain equivalent using the inverse DCT. During jpeg image compression, the DCT is used to convert 8 8blocks of the image into 64 DCT coefficients. When using the DCT coefficients in steganographic algorithms, a quantization table Q(u, v) is used to quantize the coefficients according to the formula 4 F Q (u, v) = F(u, v)/q(u, v). (1) The bits of a hidden message can then be embedded in the least significant digits of the quantized DCT coefficients. The calculations using these formulas are resource intensive. Languages such as MATLAB rich in image processing tools are used to derive the transforms. In particular, MATLAB has the DCT and Inverse Discrete Cosine Transform (IDCT) functions which can be used in processing images. There are variations of steganography algorithms that use the DCT. Some of the algorithms embed bits of a hidden message sequentially in the quantized DCT coefficients. One such algorithm is used in the JPeg-JSteg tool explained in a later section of this article. Other steganography algorithms, for example the F5 algorithm, decrements the s of the DCT coefficients using a process called matrix encoding in the embedding of a message into a stego image. h,i This use of the DCT results in a more robust algorithm; the hidden message is less likely to be detected and/or extracted. 4 Discrete Wavelet Transform The DWT is used in the JPEG 2000 compression algorithm. A wavelet is a function which integrates to 0 above and below the z-axis. Using the DWT, an image can be decomposed into wavelet coefficients. The Haar DWT is the simplest of the wavelet transforms. After performing the DWT on an image, secret information can be stored in selected DWT coefficients. 5,15 Wavelets are relatively new in image processing. They are less resource intensive and result in less image distortion than the DFT and the DCT. Wavelets are used in image processing for noise reduction, edge detection, and compression. 14 There is a wavelet toolbox in MATLAB that provides functions for processing wavelets. There are also many wavelet toolboxes available as free ware. Genetic Algorithm The GA-based steganographic system proposed by Wu and Shih 7 is also in the Frequency domain. The GA generates the stego image by artificially counterfeiting statistical features in such a way as to break the steganalytic system. The GA has its origin in natural genetics. It consists of the key concepts: reproduction, crossover, mutation, and fitness function. As applied to steganography, the GA algorithm is used to correct rounding errors that occur in processing an image transformed by DCT. After applying the DCT, embedding the secret information and then applying the IDCT to transform the image back into the spatial domain, the GA algorithm is used to translate the real numbers into integers. 5 In summary, each of the approaches described have strengths and weaknesses which make them susceptible to steganalytic techniques and applicable to specific applications and image types. STEGANALYSIS Steganalysis is the process used to detect secret information embedded in images through steganography. Most techniques used in steganography alter the characteristics and statistics of the cover image in some way. 3,16 Statistical analysis of images can detect if an image has been modified with steganography. In correspondence to the steganographic techniques, steganalysis systems fall into the same two broad categories: spatial domain steganalytic systems (SDSS) and frequency domain steganalytic systems (FDSS). SDSS is used to analyze characteristics in the spatial domain image statistics. FDSS is used to analyze characteristics in the frequency domain image statistics. 7 Detecting hidden information can be quite complex without knowing which steganalytic technique was used or if a stego key was used. But most of the steganographic techniques and tools alter the images in unique ways which serve as their signature. This helps in the detection of altered images. 11 Volume 3, May/June 2011 2011 John Wiley & Sons, Inc. 255

www.wiley.com/wires/compstats There are two main methods for detecting images modified with steganography: The first is visual analysis. Visual analysis compares the original image with the stego image either visually or using a computer to detect hidden information. Some of the simpler steganographic tools in the image domain embed information in bits without regard to the content of the carrier image. The information may be inserted in bits that make the change in the image s appearance more easily detected through visual inspection. However, this method is often not feasible because the original image is not available. The other method is through statistical analysis. 5 This method looks for anomalies in the structure or statistical measures of the image. Other steganalytic methods analyze statistical measures of an image that are considered the norm. A variation from the norm in the statistics points to an image having been altered with steganography. 3 In addition to detecting the use of steganography, a more active approach to steganalysis involves extracting and destroying the embedded hidden information. 3 It is easier to destroy messages embedded with LSB insertion than those embedded using the transforms. 11 Steganalytic techniques can vary greatly depending on what information is known about the carrier image, the stego image, the message, and the algorithm used to embed the hidden message. For example, if an image is suspected of carrying a hidden message, it may be visually inspected for irregularities and then statistical analysis through analyzing means, variances, and chi-square tests may be conducted. However, many of the steganalytic techniques used currently depend on detecting the signature of the steganographic tool used to create the stego image. 3 Such is the case with the StegDetect software explored for this article. It can be used only on jpeg images and it detects messages embedded with the known tools. For example, when StegDetect was used to analyze the image used to experiment with JSteg-JPeg for this article, it did not detect the embedded hidden message. Steganography Software Software systems have been developed to embed stego messages in images. These systems utilize some of the steganographic techniques explained. Many of the systems available are free ware and can be readily downloaded, among them JSteg, JPHide, S-Tools, and others. These systems have strengths and weaknesses which have been explored in the literature. 1,3,4,11,16 In this section, we will explore S-Tools, EzStego, and JSteg-JPeg. S-Tools is used to hide secret information in BMP, GIF, and WAV files. The messages embedded in the cover objects are encrypted using several different encryption algorithms. This tool uses LSB substitution for lossless file formats. It also uses a pseudorandom number for the LSB substitution which makes extracting the message more difficult. S-Tools can be used to embed and extract the hidden information from stego images. 5,11 EzStego is used to embed a secret message in a GIF file by modulating LSBs. EzStego compares the bits that it wants to hide with the LSB for each pixel only changing the bit if needed. 5 It is freely available as a Java application on the web. JSteg-JPeg uses the DCT transform for embedding stego messages in multiple image formats and saves the stego image as a jpeg image file. 5 Its embedding algorithm sequentially replaces the least significant digits of the DCT coefficients of the image. S-Tools S-Tools is available as free ware and can be readily downloaded from steganography sites on the Internet. The S-Tools software was developed by Andrew Brown. As an experiment for this article, messages were embedded in lena_256.bmp using S-Tools version 4. S-Tools is easy to use. It consists of a simple graphical user interface which is integrated with Windows Internet Explorer. The cover image and the file to be embedded can be dragged into the S-Tools work area from a windows explorer window. The cover image can be a WAV, GIF, or BMP file. The message to be embedded can be stored in another file such as a TXT file. The hidden information is encrypted before it is embedded in the stego image. The encryption algorithm to be used is selected from a drop-down list. A password is also required to embed the information. The password is required to extract the hidden message using S-Tools. For this experiment, text files of different sizes were embedded in Lena. One file, hide.txt, was very small and contained the phrase, Happy Birthday. Another file, starspangled.txt, contained the lyrics to the Star Spangled Banner. Before embedding text, Lena consisted of 66,614 bytes. The size of Lena did not change regardless of the amount of text stored in the image. A status message was displayed in the S-Tools work area, showing the maximum number of bytes that could be hidden in Lena (Figure 1). Extracting hidden text file was also very straight forward but required knowledge of the password and the encryption algorithm used when creating the stego image. There appeared to be a lightening in the appearance of Lena when a small text file containing the text Happy Birthday was embedded in Lena. The difference in the before and after Lena image was not noticeable after embedding the text file containing the lyrics 256 2011 John Wiley & Sons, Inc. Volume 3, May/June 2011

WIREs Computational Statistics Image steganography and steganalysis FIGURE 1 S-Tools work area. of Star Spangled Banner (Figure 2). Without having both the cover and stego image side by side there would not be any indication to the human eye that the Lena image was altered in either case (Figure 2). JPeg-JSteg JPeg-JSteg is also available as free ware and can be readily downloaded from steganography sites on the Internet. The JPeg-JSteg (JSteg) software was developed by Derek Upham. JSteg runs at the DOS prompt using the commands cjpeg and djpeg. Although JSteg outputs stego images as jpeg files, it does not read the jpeg file format. Cover images are converted using the DOS command djpeg to the Targa image file (.tga file extension). The cover image converted to the Targa image format and the hidden text file are input to JSteg using the DOS command cjpeg which produces the stego image. The algorithm used in JPeg-JSteg sequentially replaces the LSB of the DCT coefficients of the stego images with bits from the hidden message to be embedded. According to Niels Provos and Peter Honeyman, the algorithm used by JSteg can be described by the following pseudo code (Ref 4, p. 34): Msg = hidden message Pic = cover image Output = stego image While more characters in msg do: Get next DCT coefficient If DCT coefficient!= 0 and DCT coefficient!=1 Get next LSB from message Replace DCT LSB with message LSB End if Insert DCT into stego image End While The algorithm used in JPeg-JSteg inserts the steganographic secret data in the cover image during the jpeg compression algorithm after the DCT and quantization of the DCT coefficients and before the Huffmann coding step. The algorithm also embeds a length field in the stego image which is used to extract the hidden message. 10 As an experiment for this article, messages of different sizes were embedded in a jpeg image, boys.jpg. Before embedding the message, the jpeg file was converted to a tga image file using the djpeg command in DOS mode. Using the cjpeg command in DOS mode the lyrics of the Star Spangled Banner, starspangled.text, were embedded in the tga image file and then extracted. A password was not required for the embedding or extraction process. The hidden message was extracted without any problems. There appeared to be a slight lightening in the appearance of the cover image. Without having the images side by side, there was no indication that the cover image had been altered to embed a secret message (Figures 3 and 4). The size of the image remained the same. Volume 3, May/June 2011 2011 John Wiley & Sons, Inc. 257

www.wiley.com/wires/compstats Original Lena Star Spangled Banner encrypted in Lena FIGURE 2 Before and after Lena using S-Tools. FIGURE 3 Cover image used for JPeg-JSteg experiment (boys.jpg). Steganalysis Software Software systems have also been developed to detect stego messages embedded in images using steganography. Many of the steganography tools available are specific to the software used to embed the stego message. An example of such software is StegDetect. StegDetect was developed by Niels Provos. StegDetect can be used to detect jpeg images that have been altered using Steg, JPhide, Invisible Secrets, Outguess, F5, and others. The tools which StegDetect targets all use some variation of modifying bits after applying the DCT transform. 4 StegDetect uses knowledge of the technique used by the stenography software to embed the message in its detection mechanism. StegDetect can be downloaded in DOS form as free ware from the Internet. StegDetect was used to analyze the stego image created for this article (Figure 4). It gave a negative reading and was not able to detect that the image had been tampered with. FIGURE 4 Stego image produced by embedding starspangled.txt (boyschanged.jpg). CONCLUSION Steganography and steganalysis are relatively new disciplines with many relevant applications in today s digital society. The use of steganalysis is likely to increase in computer forensics in the near future. There is significant research being conducted in academic circles on steganographic and steganalytic techniques. A number of steganography tools are available on the Internet as free ware. It is not known with certainty if the use of steganography for illegal activities is widespread. There are a number of algorithms written and posted out on the Internet for anyone to download. This makes the use of steganography much easier and available for anyone who chooses to abuse the technology for illegal activities. A controlled repository of steganography and steganalysis software that can be used by students of steganography for continued 258 2011 John Wiley & Sons, Inc. Volume 3, May/June 2011

WIREs Computational Statistics Image steganography and steganalysis research. Many of the articles on steganography and steganalysis are aimed at the scientific and academic community and there is not a great deal of literature available as a starting point for the novice student of steganography and steganalysis. In conclusion, steganography and steganalysis are a growing discipline. The need for better techniques, algorithm, and software for steganalysis will continue to increase in our digital society. NOTES a Steganalyst: an analyst who uses steganalysis to detect images which have been altered using steganography. b Spatial domain: a classification applied to steganography techniques which embed the secret messages directly into the intensity s of the image pixels (example LSB insertion). c Frequency domain: a classification applied to steganography techniques which embed the secret messages after manipulating the images with transforms or algorithms. d Image domain: see spatial domain. e Transform domain: see frequency domain. f Cover image: Original image in which secret information is embedded using steganography. g Stego key: an encryption key used in addition to steganography to further secure the secret information being transmitted. h Stego image: The resulting image after secret information is embedded in a cover image using steganography. i Stego medium: see Stego image. REFERENCES 1. Johnson NF, Jajodia S. Exploring steganography: seeing the unseen, Computer 1998, 31:26 34. 2. Petitcolas FAP, Anderson RJ, Kuhn MG. Information hiding a survey. Proceedings of the IEEE, Special Issue on Protection of Multimedia Content; 1999, 1062 1078. 3. Kessler GC. An overview of steganography for the computer forensics examiner. Forensic Sci Commun 2004, 6:1 29. 4. Provos N, Honeyman P. Hide and seek: an introduction to steganography.ieee Secur Priv Mag 2003, 1:32 44. 5. Shih FY. Digital Watermarking and Steganography: Fundamentals and Techniques. Boca Raton, FL: CRC Press Inc.; 2007. 6. Provos N, Honeyman P. Detecting steganographic content on the internet. CITI Technical Report 01 11; 2001. 7. Wu YT, Shih FY. Genetic algorithm based methodology for breaking the steganalytic systems. IEEE Trans Syst Man Cybern B Cybern 2006, 36:24 31. 8. Manoj R. Understanding digital steganography, InfoSecurity Magazine, September 2009. 9. Morkel T, Eloff JHP, Olivier MS. An overview of image steganography. Proceedings of the Fifth Annual Information Security South Africa conference, (ISSA2005), Sandton, South Africa, June/July 2005. 10. Venkatraman S, Ajith A, Marcin P. Significance of steganography on data security. Proceedings of the International Conference on Information Technology: Coding and Computing (ITCC 04). ISBN:0-7696-2108-8/04. 11. Johnson NF, Jajodia S. Steganalysis of images created using current steganography software. Lect Notes Comput Sci 1998, 1525:273 289. 12. Kharrazi M, Sencar H, Memon N. Image steganography: concepts and practices. WSPC/Lecture Notes Series 2004. 13. Gonzalez RC, Woods R. Digital Image Processing. Upper Saddle Rive, NJ: Pearson Prentice Hall; 2008. 14. McAndrew A. Introduction to Digital Image Processing with MATLAB. Boston, MA: Thomson Course Technology; 2004. 15. Chen PY, Lin HJ. A DWT based approach for image steganography. Int J Appl Sci Eng 2006, 4:275 290. 16. Provos N, Honeyman P. Detecting Steganographic Content on the Internet, Center for Information Technology Integration, ISOC NDSS 02, San Diego, CA: University of Michigan; 2002. FURTHER READING Shih FY, Edupuganti VG. A differential evolution based algorithm for breaking the visual steganalytic system. Soft Comput 2009, 13:345 353. Simmons GJ. Prisoners problem and the subliminal problem. In: David Chaum, ed. Advances in Cryptology: Proceedings of Crypto 83. Springer: New York; 1998. Volume 3, May/June 2011 2011 John Wiley & Sons, Inc. 259